Search

Climate change threatens our ability to grow food for an ever-increasing population. There is a need for high-quality soil moisture predictions in under-monitored regions like Africa. However, it is unclear if soil moisture processes are globally similar enough to allow our models trained on available in situ data to maintain accuracy in unmonitored regions. We present a multitask long short-term memory (LSTM) model that learns simultaneously from global satellite-based data and in situ soil moisture data. This model is evaluated in both random spatial holdout mode and continental holdout mode (trained on some continents, tested on a different one). The model compared favorably to current land surface models, satellite products, and a candidate machine learning model, reaching a global median correlation of 0.792 for the random spatial holdout test. It behaved surprisingly well in Africa and Australia, showing high correlation even when we excluded their sites from the training set, but it performed relatively poorly in Alaska where rapid changes are occurring. In all but one continent (Asia), the multitask model in the worst-case scenario test performed better than the soil moisture active passive (SMAP) 9 km product. Factorial analysis has shown that the LSTM model's accuracy varies with terrain aspect, resulting in lower performance for dry and south-facing slopes or wet and north-facing slopes. This knowledge helps us apply the model while understanding its limitations. This model is being integrated into an operational agricultural assistance application which currently provides information to 13 million African farmers. [ABSTRACT FROM AUTHOR]

Climate change threatens our ability to grow food for a growing population. There are concurrent droughts and floods happening globally, with the greatest impacts felt in Africa. There is a need for high-quality soil moisture predictions in under-monitored regions like Africa. Yet it is unclear if soil moisture processes are globally similar enough to allow our models to maintain accuracy in unmonitored regions. We present a multitask long short-term memory (LSTM) model that learns simultaneously from global satellite-based and in-situ soil moisture data. This model is evaluated in both random spatial holdout mode and continental holdout mode (trained on some continents, tested on a different one). The model compared favorably to current land surface models, satellite products, and a candidate machine learning model, reaching a global median correlation of 0.792 for the random spatial holdout test. It behaved surprisingly well in Africa and Australia, showing high correlation even when we excluded their sites from the training set, but performed relatively poorly in Alaska where rapid changes are occurring. In all but one continent (Asia), the multitask model in the worst-case scenario test performed better than the soil moisture active passive (SMAP) 9-km product. Factorial analysis shows that the LSTM model’s accuracy with representing impacts of terrain aspect, resulting in lower performance for dry and south-facing slopes or wet and north-facing slopes. This knowledge helps us apply the model while understanding its limitations. This model is being integrated into an operational agricultural assistance application which currently provides information to 13 million African farmers. [ABSTRACT FROM AUTHOR]

Projected increases in air temperature and precipitation due to climate change in Arctic wetlands could dramatically affect ecosystem functioning. As a consequence, it is important to define the controls on evapotranspiration, which is the major pathway of water loss from these systems. We quantified the multi-year controls on midday arctic coastal wetland evapotranspiration measured with the eddy covariance method at two vegetated drained thaw lake basins near Barrow, Alaska. Variations in near-surface soil moisture and atmospheric vapor pressure deficits were found to have nonlinear effects on midday evapotranspiration rates. Vapor pressure deficits near and above 0.3 kPa appeared to be an important hydrological threshold, allowing latent heat fluxes to persistently exceed sensible heat fluxes. Dry soils increased the bulk surface resistance (water-limited). Wet soils favored ground heat flux and therefore limited the energy available to sensible and latent heat fluxes (energy-limited). Thus, midday evapotranspiration was suppressed on both dry and wet soils through different mechanisms. We also found that wet soils (ponding excluded) combined with large atmospheric demands resulted in an increased bulk surface resistance and therefore suppressing the evapotranspiration to below its potential rate (Priestley-Taylor α < 1.26). This is likely caused by the limited ability of mosses to transfer moisture during large atmospheric demands. Ultimately, in addition to net radiation, the various controlling factors on midday evapotranspiration (near-surface soil moisture, atmospheric vapor pressure, and the limited ability of mosses that are saturated at depth to transfer water during high atmospheric vapor demands) resulted in an average evapotranspiration rate of up to 75% of the potential evapotranspiration rate. The multiple limitations on midday evapotranspiration rates have the potential to moderate interannual variation of total evapotranspiration and dampen excessive water loss during a warmer climate. Combined with the prevailing maritime winds and the projected increase in precipitation, these dampening mechanisms will likely prevent extensive future soil drying and hence maintain the presence of coastal wetlands. [ABSTRACT FROM AUTHOR]